This is a project for the Machine Learning Course by Prof.Luca Iocchi and Prof.Valsamis Ntouskos @ Sapienza University di Roma
Maritime Detection, Classification and tracking data set and can be downloaded from http://www.dis.uniroma1.it/~labrococo/MAR/classification.htm The training set contains images from 24 different categories of boats navigating in the City of Venice (Italy). The .rar file contains a folder for each category. The jpeg files inside the folders are named according to the date, hour, and system track number. The folder "Water" contains false positives
1.Tensorflow
https://www.tensorflow.org/install/install_linux
2.Pandas
pip install pandas
3.Numpy
pip install numpy
4.clone this repo
git clone git@github.com:adiltirur1/Image_classification.git
Copy the training and testing data set in to
/image_classifier/supporting_files/data_set
from the root directory of this repository run the command
python -m code.retrain --bottleneck_dir=supporting_files/bottlenecks --how_many_training_steps=500 --model_dir=supporting_files/models/ --summaries_dir=supporting_files/training_summaries/"${ARCHITECTURE}" --output_graph=supporting_files/retrained_graph.pb --output_labels=supporting_files/retrained_labels.txt --inception_v3 --image_dir=supporting_files/sc5
from the root directory again run the following command
from the root directory again run the following command
python -m code.bulk_classify --graph=supporting_files/retrained_graph.pb
As an output of the above program, there will be a csv file created image_classification/supporting_files/results/pre_result.csv
This csv file contain the predicted output and the actual output according to the ground truth seperated by ','
navigate to
image_classification/code
Run the command
python accuracy.py